Journal of

Mechanics of
Materials and Structures

TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION

BEARINGS
James M. Kelly and Shakhzod M. Takhirov

Volume 2, Nº 8 October 2007

mathematical sciences publishers
 JOURNAL OF MECHANICS OF MATERIALS AND STRUCTURES
Vol. 2, No. 8, 2007

TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS

JAMES M. K ELLY AND S HAKHZOD M. TAKHIROV

Seismic isolators are constructed from multiple layers of elastomer (usually natural rubber) reinforced
with steel plates; they are, therefore, very stiff in the vertical direction, but soft in the horizontal direction.
The buckling of these bearings under compression load is a well-understood phenomenon and has been
widely studied. It is therefore unexpected that the buckling analysis for compression predicts that the
isolator can buckle in tension at a load close to that for buckling in compression. The linear elastic model
that leads to both compression and tension buckling is an extremely simple one, so it might be argued
that the tensile buckling may be an artifact of the model itself rather than a property of the isolator. To
test the simple theoretical model we have conducted a numerical simulation study using a finite element
model of a multilayer elastomeric bearing. We find that the prediction of tensile buckling by the simple
linear elastic theory is indeed accurate and not an artifact of the model.

1. Introduction

Seismic isolation using multilayer elastomeric isolators has been used in the United States and around the
world for more than 20 years. The isolators are constructed from many layers of elastomer reinforced
with steel shims, and are very stiff in the vertical direction, but soft in the horizontal direction. This
enables them to carry the weight of a building, but cause the building to have a fundamental natural
frequency that is both lower than that of the same building, if conventionally founded, and the dominant
frequencies of strong ground motion.
They appear to be very stable, though the low shear stiffness causes a buckling phenomenon. However,
it is straightforward to design them to have a large safety factor against buckling. The buckling of
these bearings under compression load is a well-understood phenomenon, and has been widely studied.
Buckling theory is based on a linearly elastic analysis. Although the elastomer is not really linearly
elastic, the deformation is predominantly one of shear, and typical elastomers used in bearings are very
close to linear over a large range of shear strain. While approximate, the linear theory is relatively
accurate and adequate for most design purposes.
However, buckling analysis for compression has been used to make an unexpected prediction that the
isolator can buckle in tension at a load close to that for buckling in compression. Of course, there are
many examples of strange systems that buckle in tension, but these are entirely pathological in that the
tension forces are always transferred to compression elements that produce the instability. This is not the
case here. The buckling process is in fact tensile. The linear elastic model that leads to both compression
and tension buckling is an extremely simple one, so it might be argued that tensile buckling may be an
artifact of the model itself and not of the isolator.
Keywords: steel-reinforced elastomeric seismic isolators, tension and compression buckling, linear theory, nonlinear finite
element analysis.

1591
1592 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

For this reason we had undertaken a numerical simulation study using a finite element model of a
multilayer elastomeric bearing to check whether the prediction of tensile buckling by the simple linear
elastic theory is, in fact, accurate. We found this to be the case. The essential point is that the mechanics
of the isolator in tension are the mirror image of that for the isolator in compression. In particular, when
the isolator is in compression below the buckling load but laterally displaced, the layers in the center
experience rotations that give the vertical load a component along the layer causing a shear deformation.
In tension, the layers in the center experience rotations in the opposite direction giving a shear defor-
mation due to the tensile force that permits the top of the isolator to move upwards by a much larger
displacement than that which could be sustained in pure tension with no lateral displacement.
While the bearings are normally in compression, base-isolated tall buildings in near-fault locations
can lead to situations where peripheral bearings in the isolation system can be required to take some
amount of tension. This tension is caused by global overturning of the building produced by the lateral
inertial force at the center of the mass of the isolated building. The maximum inertial force and the
resulting maximum overturning movement occur at the same time as the maximum lateral displacement
of the isolators; at first sight this would seem to be a critical situation. The value of the buckling analysis
is that it demonstrates that the condition of the isolator in tension and shear is not as dire as had been
feared. In tension, the layers in the center experience a rotation which allows a shear deformation caused
by the tensile force and permits the top of the isolator to move upwards by a much larger displacement
than that which could be sustained in pure tension with no lateral displacement. Thus the simultaneous
occurrence of tension and shear in the isolator prevents the development of damage due to cavitation.

2. Overview

Earliest theoretical approaches to study the stability of rubber bearings by Haringx [1948; 1949a; 1949b]
were based on linearity of the rubber material and small displacements. Theoretical predictions of the
decrease in horizontal stiffness with increasing axial load based on Haringx’s theory were made by
Gent [1964] and Derham and Thomas [1981]. Simo and Kelly [1984] used finite element modeling to
study the variation of lateral load-displacement behavior under increasing axial load.
An extensive experimental study of low shape factor elastomeric isolators used for base isolation was
given by Aiken et al. [1989]. Buckling tests were conducted on doweled bearings; they consisted of
applying monotonically increasing axial load to a bearing with the top of the bearing free to displace in
the horizontal direction. The tests showed that the analytical formula gives a higher value of the critical
load than the experimentally measured one.
Roeder et al. [1987] and Stanton et al. [1990] studied the stability of laminated elastomeric bearings ex-
perimentally and theoretically with due consideration given to axial shortening. Buckle and Kelly [1986]
studied the stability of elastomeric bearings using a model bridge deck tested using a shaking table.
Since the bearings were doweled, bearing overturning or rollover was clearly evident in these tests.
Koh and Kelly [1986; 1988; 1989] developed a viscoelastic stability model and a mechanical model
based on bearing test results. A comprehensive study of the basic theory and its application to design
issues, including the stability problem, was presented by Kelly [1997].
Experimental determination of critical buckling behavior of steel-reinforced bearings at high shear
strain was conducted by Buckle and Liu [1994]. Studies by Nagarajaiah and Ferrell [1999] introduced
 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1593

a nonlinear analytical model based on the Koh–Kelly model and included large displacements, large
rotations, and nonlinearity of the rubber. The model was verified through experimental results. Further
experimental work on the stability of elastomeric bearings was presented by Buckle et al. [2002]. It was
shown that the critical buckling load decreases with increasing horizontal displacement or shear strain.
Finite element analysis results conducted on a plane model of the bearings with a coarse mesh were
compared with the experimental results.
Some results on finite element analysis of the multilayered elastomeric bearings can be found in several
papers and reports; see, for example, [Takayama et al. 1992].

3. Formulation of elementary stability theory of multilayer elastomeric isolators

The elementary theory for the buckling of isolation bearings treats the bearing as a continuous homo-
geneous beam in which plane sections normal to the undeformed axis remain plane but not necessarily
normal to the deformed axis. The deformation is defined by three functions u(x), v(x), ψ(x), which
are the axial and lateral displacements of the centroidal axis and the rotation of a section normal to
the undeformed axis, respectively. The overall shear deformation γ (x) of the section is the difference
between the rotations of the centroidal axis and the section, namely, γ (x) = v 0 (x) − ψ(x).
The internal forces on the deformed plane section are the axial load N (x) normal to the section, the
shear force V (x) parallel to the section and the bending moment M(x), as shown in Figure 1. These
internal forces are related to the deformation quantities through

N (x) = E c As u 0 (x), V (x) = G As (v 0 − ψ), M(x) = E Is ψ 0 (x).

As is the cross sectional area A increased by h/tr , where h is the total height of the bearing (rubber plus
steel), tr is the total thickness of rubber. and E c is the compression modulus of the bearing. The value of
E c for a single rubber layer is controlled by the shape factor S =(loaded area)/(force-free area), which
is a dimensionless aspect ratio of a single layer of the elastomer. For example,

 S = b/t,
 infinite strip of width 2b with a single layer thickness t,
S = S = R/2t, circular pad of diameter R and thickness t,


S = a/4t, square of side a and thickness t.

In a circular isolator E c = 6G S 2 , while in the long strip isolator that will be used for the numerical
simulation E c = 4G S 2 . The increase in the area is needed to account for the fact that the steel does not
deform in the composite beam. Is is the effective moment of inertia of the cross section. This is modified
the same way as A. There is an additional modification to account for the fact that in the isolator the
pressure distribution that generates the internal bending moment is a cubic parabola in contrast to regular
beam theory where the bending stress distribution is linear. The effective bending stiffness, accounting
for these two effects is [Kelly 1997]
(
1
E c I thr , circular bearing,
E Is = 13
5 E c I tr , strip bearing.
h
1594 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

Under the kinematic assumptions of this beam theory, the downward deflection of the top of the
composite column representing the isolator and the resulting external work done by the applied load P
are, respectively,
Z h
P h
Z
v(x) = 2
1
2v ψ − ψ d x,
0 2
2v 0 ψ − ψ 2 d x.



Wext = − Pv(x) = −
0 2 0
The internal stored energy of the column is given by
Z h Z Z
1
N (x)u d x + 2 M(x)ψ (x)d x + 2 V (x) v 0 (x) − ψ(x) d x.
0 1 0 1


Wint = 2
0

Application of the method of virtual work to the total work Wint + Wext with respect to the virtual
displacements δu(x), δv(x), δψ(x) leads to the following set of equilibrium equations
N 0 = 0, V 0 − Pψ = 0, M 0 + V + P(v 0 − ψ) = 0. (1)
The boundary conditions for Equations (1) are shown in Figure 1; the external loads applied to the column
at x = 0 are the axial load P (or T ), a transverse load Ho and a moment Mo .
From Equation (1)1 we have N = const = − P. Integrating the second equation using the consistent
boundary condition gives
V − Pψ = − Ho , (2)
and inserting V = Pψ − Ho into the third equation gives M 0 + Pv 0 = Ho , which can be integrated to
M + Pv = Mo + Pvo + Ho x, (3)
where vo = v(0). These two integrated versions of equilibrium equations (2) and (3) are the starting point
for the stability analysis of the isolator. When the constitutive equations are included, these become
E Is ψ 0 + Pv = Pvo + Mo + Ho x, G As (v 0 − ψ) − Pψ = − Ho .

Figure 1. Geometry and loads in theoretical study of buckling.

 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1595

The second equation above can be used in two ways to give a pair of uncoupled equations for v and ψ.
First we can write ψ in terms of v in the form ψ = (G As v 0 + Ho )/(G As + P), and substitute ψ 0 into the
first equation to get
G As
E Is v 00 + Pv = Pvo + Mo + Ho x. (4)
G As + P
We can also express v in terms of ψ as v 0 = (P + G As )ψ/(G As ) − (Ho )/(G As ), which, when substituted
into the first, leads to
G As
E Is ψ 00 + Pψ = Ho . (5)
G As + P
Thus, Equations (4) and (5) for two kinematic variables have similar form, but different right hand sides.
The most general form of solution, taking into account the connections between v and ψ, is
Mo Ho Ho
v = A cos αx + B sin αx + vo + + x, ψ = αβ B cos αx − αβ A sin αx + ,
P P P
where A and B are constants of integration and α, β are given by
P(P + G As ) G As
α2 = , β= .
E Is G A s P + G As
These equations are now used to predict the buckling behavior of a bearing in an isolation system. As
shown in Figure 2, the isolator is constrained against displacement and rotation at the bottom, against
rotation at the top, but is free to move laterally at the top, which give boundary conditions v(0) = 0,
ψ(0) = 0, ψ(0) = 0, Ho = 0 leading to αh = π. The deformed configuration becomes
 πx  πx
v(x) = 12 δh 1 − cos , ψ(x) = 21 αβδh sin .
h h

Figure 2. Boundary conditions for isolation bearing under a vertical load P. The bear-
ing buckles with no lateral force constraint, but is prevented from rotating at each end.
1596 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

The above result α 2 = π 2 / h 2 means that

P(P + G As ) π 2 E Is
= = PE ,
G As h2
where PE is the Euler load for the column without shear deformation. Denoting G As by PS , the above
equation for the buckling load can be recast as a quadratic P 2 + P PS − PS PE = 0, which gives two
critical loads: a compression load PC and a tension load PT
q q
−PS + PS2 + 4PS PE −PS − PS2 + 4PS PE
PC = , PT = . (6)
2 2
For all reasonable values of the shape factor S, the PS is so much less than PE that it can be neglected,
and the two critical loads can be approximated by
p
PC,T = ± PS PE . (7)

The significance of the tensile critical load becomes clear when we replace the generic load P in the
formulae for α and β by −T , with the assumption T ≥ G As , giving
T (T − G As ) G As
α2 = , β= − ,
E Is G A s T − G As
showing that the buckled shape in tension v(x) is the same as that in compression, but the rotation ψ(x)
is reversed, and the central layers of the bearing are rotated in the direction that facilitates the upward
movement of the top through a rotated shear deformation. Because of the natural symmetry of shear,
there is an intrinsic symmetry here between compression and tension.

4. Modification for change in length of column prior to buckling

One interesting aspect of the comparison between the buckling loads predicted by the theory and those
obtained by the numerical simulation is the fact that while the theory has the buckling load in tension
always slightly higher than that in compression for the same shape factor (the difference is G As ), in the
numerical results the buckling load in compression is always higher than that in tension. The reason for
this is that theory neglects the change in length due to the axial load, whereas in the simulation when
buckling is initiated, the bearing has shortened or lengthened. For smaller values of the shape factor, the
changes in length can be significant.
Using the approximation G As  P and the values of G As , E Is for a long strip √ bearing, the critical
pressures pcrit = P/A are given in the theoretical analysis by pcrit /G = ±2πbS/( 15tr ), where tr is the
total thickness of rubber in the bearing. To compare the theory with the simulation we replace tr by
 pcrit 
tr = tro 1 ∓ ,
4G S 2
where the minus sign is for compression and the plus for tension. The buckling loads are then given by
pcrit 2πbS
G
= ±√  p .
15tro 1 ∓ crit 2
4G S
 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1597

24
Approximation
22 PC
PT
20
P*C

Critical load normalized by GA

18 P*T

16

14

12

10

4
2 4 6 8 10 12 14
Shape factor

Figure 3. Theoretical buckling loads with and without global bearing deformation:
Equation (6) versus (8); solid line corresponds to load approximation from (7).

Solving the last expression for pcrit /G and taking into account that P = p A leads to
( s )
∗
PC,T 2π b
2
= ±2S 1 − 1 ∓ √ . (8)
GA Str0 15
Compression and tension critical loads computed by means of Equation (8) are presented in Figure 3.
It is easy to see that the value in compression is always larger than that in tension. The difference
between |Pcrit /G A| in compression and tension is approximately 2(2π 2 b2 )/(15tro 2 ). For the case of the
numerical models studied we have b = tro . The difference is about 8/3 and, thus, becomes less important
with increasing S.

5. Numerical modeling of buckling in tension

In order to verify the tension buckling predicted by the simple analytical theory we use the general
purpose finite element ABAQUS application [ABAQUS 2001] to model a steel-reinforced bearing and
to study the buckling behavior in both tension and compression [Kelly and Takhirov 2004].

5.1. Modeling details. Five finite element models of a bearing are created. These numerical models
have the same width and total rubber thickness, and steel shims have the same thickness also. The
only difference between them is the shape factor of the bearing. The total thickness of rubber layers
is the same in each model, but the thickness of a single rubber layer varies from model to model. The
correspondence between the model name and the shim thickness is given in Table 1.
The finite element analysis on the elastomeric bearings is restricted to plane strain. The O y-axis of
the coordinate system is a vertical axis that extends across the steel shims and rubber layers, while the
1598 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

Steel shim Total rubber Rubber layer

Model thickness thickness Width thickness Shape factor

Model 1 2.60 80.01 160.02 5.72 14

Model 2 2.60 80.01 160.02 8.00 10
Model 3 2.60 80.01 160.02 11.43 7
Model 4 2.60 80.01 160.02 16.00 5
Model 5 2.60 80.01 160.02 26.67 3

Table 1. Numerical models with various shape factors. All lengths in mm.

horizontal axis O x corresponds to the lateral direction of the bearing as shown in Figure 4. Generally
steel-reinforced rubber bearings have a hole in the middle of the steel plates and a rubber cover on the
traction-free sides of the bearing. In order to create a model close to the theoretical one given earlier, the
hole in the middle and the rubber cover are not included in the consideration.
In the analysis the end plates of the bearing are assumed to be undeformable. Therefore, in the
numerical model, the top rubber layer of the bearing is connected to an absolutely rigid surface with the
reference point in the middle at which the vertical load is applied. The bottom surface of the bearing
is fixed. Two vertical sides of the bearing model are traction-free. The top surface is restrained against
rotation around the Oz-axis (out of the plane), but is free to move horizontally.
Linearly elastic material properties are assumed for the steel plates with Young’s modulus and Pois-
son’s ratio equal to 200, 000 MPa and 0.3, respectively. Rubber materials have very little compressibility
compared to their shear flexibility; these materials are usually modeled by a hyperelastic material model.
ABAQUS [ABAQUS 2001] has a special family of hybrid elements to model the fully incompressible
behavior seen in a rubber material. The following assumptions are made in modeling a rubber material:
(1) elastic, (2) isotropic, (3) (almost) incompressible, and (4) includes nonlinear geometric effects.

Figure 4. Geometry and coordinate axes of numerical simulation models.

 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1599

Rubber model C10 C01 C20 C11 C02 D1

Polynomial 193.4 −0.1 −0.8 0.2 0 0
neo-Hookean 345.0 0 0 0 0 9.7 × 10−7

Table 2. Material parameters for two rubber models. Ci j in kPa, D1 in kPa−1 .

Hyperelastic materials are described in terms of a strain energy potential U , which defines the strain
energy stored in the material per unit of reference volume in the initial configuration as a function of the
strain at that point in the material. The rubber is selected as a polynomial hyperelastic material of the
second order. In this case, the strain energy potential has the form
2
X (J el − 1)2
U= Ci j (I1 − 3)i (I2 − 3) j + ,
D1
i+ j=1

where Ci j and D1 are the material parameters, I1 and I2 are the first and the second invariants of the
deviatoric strain, and J el is the elastic volume ratio.
Two rubber models are included in the consideration: a fully incompressible (polynomial) model, and
an almost incompressible (neo-Hookean) model. The material parameters of the rubber can be expressed
in terms of initial shear modulus G, and initial bulk modulus K via G = 2(C10 + C01 ), K = 2/D1 . The
values of the material parameters for both rubber models are presented in Table 2. Since D1 is not equal
to zero for the neo-Hookean model, this model allows some compressibility in the rubber material.
A supplemental study on the properties of the rubber models is conducted on a rubber cylinder and a
rubber layer. The cylinder is used for the rubber material study in tension and compression. The layer,
representing one single layer of the rubber locked between two rigid horizontal surfaces, is used to study
behavior of the rubber material in shear with no vertical load. While both rubber materials are linearly
elastic up to about 250% strain in shear, they exhibit a significant nonlinearity in tension or compression
as shown in Figure 5.

5.2. Critical buckling load in compression and tension. The model is studied by a classical buckling
analysis scheme available in ABAQUS. First the buckling mode of each bearing model is determined.
Very small imperfections of about 1% of the steel layer thickness are introduced in the model; they
are based on the buckling mode obtained in the buckling mode analysis. The postbuckling behavior is
followed up to about 30% shear deformation.
The buckling analysis of the numerical bearing models reveals the following results. All models
have significant horizontal drift caused by a large vertical load. The curves of the compression vertical
load versus horizontal drift for all numerical models are shown in Figure 6. The critical buckling load
increases with the increase in the shape factor. All plots show similar behavior with the majority of the
change happening up to a 3% deformation after which they flatten out when the bearing is buckling.
Figure 7 presents the corresponding curves of buckling in tension. The critical load is again dependent
on the shape factor increasing with it. The buckling in tension is more sudden, so the point at which
buckling begins moves very close to the vertical axis, and the shear when the buckling starts in tension
1600 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

3
FEA: Polynomial
Linear theory
2.5

Shear stress normalized by G

2

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3
Shear strain, in/in
4
True stress
2 Nominal stress
Compression/tension stress normalized by G

Linear theory
0

−2

−4

−6

−8

−10

−12

−14

−16
−0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1
Longitudinal strain, in/in

Figure 5. Properties of rubber material in (top) shear and (bottom) compression/tension

deformation (polynomial rubber material).

can be as low as 0.2%; see Model 5 with the smallest shape factor. Increase in the shape factor moves
this point closer to the 3% critical strain obtained for the compression buckling.
Theoretical critical force versus the shape factor for the corresponding numerical model was discussed
earlier; see Figure 3. Theoretical buckling compression load is always less than the absolute value of the
tension load for the simple theoretical solution presented in Equation (6), which is not consistent with
the numerical analysis results shown in Figure 8. The critical load determined by Equation (8) takes into
account shortening or elongation of the bearing in the vertical direction. Therefore, it correlates better
with numerical results. As the latter show, the compression buckling load is always greater than the
absolute value of the tension buckling load shown in Figure 8 by dashed and dot-dashed lines. Figure 8
 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1601

25

20

Compression load normalized by GA

15

10

S = 14
S = 10
5
S=7
S=5
S=3
0
0 5 10 15 20 25 30 35
Horizontal deformation, %

Figure 6. Buckling diagram for all models (compression).

20
S = 14
18 S = 10
S=7
16
S=5
Tension load normalized by GA

S=3
14

12

10

0
0 5 10 15 20 25 30 35
Horizontal deformation, %

Figure 7. Buckling diagram for all models (tension).

also shows no significant differences between critical buckling load for incompressible (polynomial) and
compressible (neo-Hookean) rubber materials. As a typical result, Figures 9–12 show deformed shapes
of two numerical models in compression and tension.
The numerical results on buckling behavior of the bearing have satisfactory correlation with the the-
oretical solutions presented earlier. The theoretical study and the finite element analysis lead to the
following conclusions. The critical buckling load increases with the shape factor; this behavior is almost
1602 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

25
Approximation
PolynomialC
PolynomialT
20 Neo−HookeanC
Neo−HookeanT

Critical load normalized by GA

P*C
P*T
S, Mises 15
(Ave. Crit.: 75%)
+2.089e+04
+1.915e+04
+1.741e+04
+1.567e+04 10
+1.393e+04
+1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03 5
+5.226e+03
+3.485e+03
+1.744e+03
+2.670e+00
0
2 4 6 8 10 12 14
Shape factor

Figure 8. Critical buckling load (normalized by GA) versus shape factor.

S, Mises
(Ave. Crit.: 75%)
+2.089e+04
+1.915e+04
+1.741e+04
+1.567e+04
+1.393e+04
+1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03
+5.226e+03
+3.485e+03
+1.744e+03
+2.670e+00 S, Mises
(Ave. Crit.: 75%)
+2.089e+04
+1.915e+04
+1.741e+04
+1.567e+04
+1.393e+04
+1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03
2 +5.226e+03
+3.485e+03
+1.744e+03
+2.670e+00
3 1 2

3 1

Figure 9. Buckled shape and Mises stresses for Model 2 (compression).

linear as predicted by the theory. The numerical compression buckling load is almost always higher than
the theoretically estimated one for all bearings. The compression buckling load is always higher than the
absolute value of the corresponding tension load due to nonlinear geometry effects noted for the simple
rubber cylinder model. The numerical models have different postbuckling behaviors in compression and
tension. In compression, the vertical load remains almost the same after the buckling occurs, and the
bearing deflects horizontally. In contrast, the vertical load in tension slowly decreases with horizontal
deflection.
2

3 1
 +1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03
+5.226e+03
+3.485e+03
+1.744e+03
+2.670e+00

TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1603

S, Mises
(Ave. Crit.: 75%)
+2.089e+04
+1.915e+04
+1.741e+04
+1.567e+04
+1.393e+04
+1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03
+5.226e+03
+3.485e+03
+1.744e+03
+2.670e+00

3 1

Figure 10. Buckled shape and Mises stresses for Model 2 (tension).
S, Mises
(Ave. Crit.: 75%)
+2.089e+04
+1.915e+04
+1.741e+04
+1.567e+04
+1.393e+04
+1.219e+04
+1.045e+04
+8.708e+03
+6.967e+03
+5.226e+03
+3.485e+03
2 +1.744e+03
+2.670e+00

3 1

Figure 11. Buckled shape and Mises stresses for Model 5 (compression).

3 1

Figure 12. Buckled shape and Mises stresses for Model 5 (tension).
1604 JAMES M. KELLY AND SHAKHZOD M. TAKHIROV

6. Conclusions

We have shown that numerical analysis confirms that the buckling of a bearing in tension is not an artifact
of the way the theory has been set up. Of course, it must be acknowledged that a bearing will experience
cavitation before the buckling load can be achieved. However, the theory shows that under high seismic
loading, the isolators that experience tension will avoid the possibility of cavitation since the tension
loading due to the overturning of the structure is accompanied by large lateral shear. Moreover, due to
the interaction between shear and the vertical stiffness, the tension stresses are much less than they would
be if the tension force were applied in the absence of shear.
It has been the purpose of this paper to demonstrate that the condition of the isolator in tension and
shear is not as dire as has been feared. The analysis has shown that the mechanics of the isolator in
tension are the mirror image of those for the isolator in compression. In particular, when the isolator
is in compression below the buckling load but laterally displaced, the layers in the center experience a
rotation which gives the vertical load a component along the layer and turns the compression displacement
into a shear deformation. In tension the rubber layers in the center of the bearing experience a rotation
in the opposite direction which allows a shear deformation caused by the tensile force and permits the
top of the isolator to move upwards by a much larger displacement than that which could be sustained
in pure tension with no lateral displacement. The elastomer can sustain only small strains in the state of
triaxial stress generated by pure tension on a multilayer isolator with a large shape factor, but can sustain
shear strains on the order of 500–600%. Thus the simultaneous occurrence of tension and shear allows
the isolator to avoid the damaging effects of cavitation.

References
[ABAQUS 2001] ABAQUS 6.2 user manual, Version 6.2, Hibbitt Karlsson Sorensen Inc., Pawtucket, RI, 2001.
[Aiken et al. 1989] I. D. Aiken, J. M. Kelly, and F. F. Tajirian, “Mechanics of low shape factor elastomeric seismic isolation
bearings”, Technical report UCB/EERC-89/13, Earthquake Engineering Research Center, University of California, Berkeley,
CA, 1989.
[Buckle and Kelly 1986] I. G. Buckle and J. M. Kelly, “Properties of slender elastomeric isolation bearings during shake table
studies of a large-scale model bridge deck”, pp. 247–269 in Joint sealing and bearing systems for concrete structures, vol. 1,
American Concrete Institute, Detroit, MI, 1986.
[Buckle and Liu 1994] I. G. Buckle and H. Liu, “Experimental determination of critical loads of elastomeric isolators at high
shear strain”, NCEER Bull. 8:3 (1994), 1–5.
[Buckle et al. 2002] I. Buckle, S. Nagarajaiah, and K. Ferrell, “Stability of elastomeric isolation bearings: experimental study”,
J. Struct. Eng. ASCE 128:1 (2002), 3–11.
[Derham and Thomas 1981] C. J. Derham and A. G. Thomas, “The design of seismic isolation bearings”, pp. 21–36 in Con-
trol of seismic response of piping systems and other structures by base isolation, Earthquake Engineering Research Center,
University of California, Berkeley, CA, 1981.
[Gent 1964] A. N. Gent, “Elastic stability of rubber compression springs”, J. Mech. Eng. Sci. 6:4 (1964), 318–326.
[Haringx 1948] J. A. Haringx, “On highly compressible helical springs and rubber rods and their application for vibration-free
mountings, I”, Philips Res. Rep. 3 (1948), 401–449.
[Haringx 1949a] J. A. Haringx, “On highly compressible helical springs and rubber rods and their application for vibration-free
mountings, II”, Philips Res. Rep. 4 (1949), 49–80.
[Haringx 1949b] J. A. Haringx, “On highly compressible helical springs and rubber rods and their application for vibration-free
mountings. III”, Philips Res. Rep. 4 (1949), 206–220.
 TENSION BUCKLING IN MULTILAYER ELASTOMERIC ISOLATION BEARINGS 1605

[Kelly 1997] J. M. Kelly, Earthquake-resistant design with rubber, 2nd ed., Springer, London, 1997.
[Kelly and Takhirov 2004] J. M. Kelly and S. M. Takhirov, “Analytical and numerical study on buckling of elastomeric bearings
with various shape factors”, Technical report UCB/EERC-2004/03, Earthquake Engineering Research Center, University of
California, Berkeley, CA, 2004.
[Koh and Kelly 1986] C. G. Koh and J. M. Kelly, “Effects of axial load on elastomeric bearings”, Technical report UCB/EERC-
86/12, Earthquake Engineering Research Center, University of California, Berkeley, CA, 1986.
[Koh and Kelly 1988] C. G. Koh and J. M. Kelly, “A simple mechanical model for elastomeric bearings used in base isolation”,
Int. J. Mech. Sci. 30:12 (1988), 933–943.
[Koh and Kelly 1989] C. G. Koh and J. M. Kelly, “Viscoelastic stability model for elastomeric isolation bearings”, J. Struct.
Eng. ASCE 115:2 (1989), 285–302.
[Nagarajaiah and Ferrell 1999] S. Nagarajaiah and K. Ferrell, “Stability of elastomeric seismic isolation bearings”, J. Struct.
Eng. ASCE 125:9 (1999), 946–954.
[Roeder et al. 1987] C. W. Roeder, J. F. Stanton, and A. W. Taylor, “Performance of elastomeric bearings”, NCHRP Rep. 298,
Trans. Res. Board, National Research Council, Washington, D. C., 1987.
[Simo and Kelly 1984] J. C. Simo and J. M. Kelly, “Finite element analysis of the stability of multilayer elastomeric bearings”,
Eng. Struct. 6:3 (1984), 162–174.
[Stanton et al. 1990] J. F. Stanton, G. Scroggins, A. W. Taylor, and C. W. Roeder, “Stability of laminated elastomeric bearings”,
J. Eng. Mech. ASCE 116:6 (1990), 1351–1371.
[Takayama et al. 1992] M. Takayama, H. Tada, and R. Tanaka, “Finite-element analysis of laminated rubber bearing used in
base-isolation system”, Rubber Chem. Technol. 65:1 (1992), 46–62. Rubber Division, ACS.

Received 31 Jul 2006. Accepted 20 Apr 2007.

JAMES M. K ELLY: jmkelly@berkeley.edu
Earthquake Engineering Research Center, 1301 South 46th St., Bldg. 451, University of California, Berkeley,
Richmond, CA 94804, United States
S HAKHZOD M. TAKHIROV: takhirov@berkeley.edu
Earthquake Engineering Research Center, 1301 South 46th St., Bldg. 451, University of California, Berkeley,
Richmond, CA 94804, United States